help-gplusplus
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: What should the call stack look like when calling a C++ virtual func


From: Dev1024
Subject: Re: What should the call stack look like when calling a C++ virtual function?
Date: 24 Sep 2006 20:22:35 -0700
User-agent: G2/1.0

A sincere thanks for your response and clarification.  In response to
why, yes, I have cases where g++ code will call MSVC code (or vice
versa).  Currently, as explained below, it doesn't work due to the this
pointer on the stack.  I tried changing the calling convention (cdecl,
stdcall, fastcall) to no avail.  Any other ideas?

Thanks,
Dev1024



Paul Pluzhnikov wrote:
> "Dev1024" <lmemmott@yahoo.com> writes:
>
> > Having been a long time user of Microsoft Visual C++ .Net 2003 (MSVC),
> > I've recently been using g++ on a project.  I've found a difference
> > between the two compilers when calling a virtual function that has me
> > puzzled.
>
> The immediate question is "why do you care?"
>
> > The issue is when calling virtual functions.  In MSVC the parameters
> > are pushed on to the stack as expected and then the address of the
> > function is loaded into EAX.  Finally "call EAX" is issues to call the
> > virtual function.
>
> That is incorrect: you are missing a step: loading "this" into %esi.
>
> > In g++ (Cygwin g++ v3.4.4) the parameters are pushed
> > onto the stack as expected, but then a pointer (32 bit pointer in my
> > case) to the vtable is pushed on the stack as well
>
> That is also incorrect: it's the "this" parameter that is pushed.
> However, since g++ puts the vtable at the start of any class with
> virtual functions, the "this" and &vtbl are the same thing.
>
> > - Why does g++ load a pointer to the vtable on the stack after pushing
> > on the params?
>
> It doesn't. It simply passes all parameters (including "this")
> on the stack.
>
> > - Is there a compiler option that can be used to avoid putting a
> > pointer to the vtable on the stack?  (I looked but couldn't find one.)
>
> These are implementation details, best left alone.
> Are you trying to link together MSVC and g++ compiled code?
>
> Cheers,
> --
> In order to understand recursion you must first understand recursion.
> Remove /-nsp/ for email.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]